How to Dry SF6 Tank: Rules of Thumb & Models

[Hypersphere]

Hello,
Now that I have an actual problem (in summer work) I thought I'd take the step from occasional forum lurking to actually post. A customer have recently had some problems with moisture in a tank of insulating SF₆ gas, protecting some 300kV electronics. Allow me to outline the construction.

The tank is cylinder shaped, 1m^3 in volume, generally filled with sulphur hexaflouride at ca 1 bar. The high voltage feedthrough from the transformer is placed in the center and the electronics compartment just above - since it needs to do some work on the HV. Now, this isn't the problematic part. In the bottom of the tank there's a small cooling unit. A fan pushes gas into a short tube, where the gas meets a heat exchanger, fed from the outside with water of temperature T, and then proceeds out through a hose pointing at the electronics compartment.

In most (non-tropic) installations T=22^\circC and then the following procedure works well for gas exchange (for maintenance of the electronics or just because discharges builds up contamination):

1. Pump to 1-10mbar
2. Nitrogen, 1 bar
3. Pump to 1-10 mbar
4. Nitrogen, 1 bar
5. Pump to 1-10mbar
6. SF₆, 1 bar

That this procedure works well is no surprise - the amount of vapor should be effectively reduced both by the pumping and by the gas replacement. Problems do however occur when the cooling water has lower temperatures and condensation occurs. For example, one installation has been "overclocked" by setting T=5^\circC, which obviously meant quite a bit of condensed water after a few years of operation. My question is now, for a given amount of condensed water (probably a fraction of which is chemisorbed), how long would I have to pump at a given pressure?

I would be grateful for known rules of thumb, or simple models. The physics person in me tried to solve this using Langmuir evaporation, but the results are unreasonably optimistic compared to what apparently did not work. Thanks in advance for your ideas.

EDIT: Going only by Langmuir is of course simplified, but I don't know how to tackle the pressure gradients etc. Somehow I feel that experience is more valuable in this case than exact answers, especially given that I don't think we want to dwell too long on this. That's also why I put in the engineering section, hopefully that reasoning is sound.

 

[eachus]

Interesting problem, but I think you are looking at it from the wrong end of the telescope. Try stating the problem thus:

I have a tank of known volume containing an unknown amount of water. I need a procedure which will dry the tank so it can be refilled with sulfur hexafluoride.

Now the first two things that come to mind are measuring either the volume of contaminants and water in the tank, or checking the humidity. And you might want to run the cooling water at room temperature or above to speed up the process.

Oh, and again with my engineer hat on, I assume you are checking for leaks while pumping down? If you stop pumping and the pressure starts rising, you either have ice in the container, or a leak.

Anyway, my back of the envelope approach would be to run the cooling water while pumping down, and set the pressure limit on the pumping based on the temperature of the cooling water. Use the vapor pressure of water as a guide, though. In a real system that water is going to have plenty of dissolved material.

Last but not least, have you thought about what concentrated acid may do to the container and its contents? The contaminants are going to include some HF, H2S, H2SO3, and H2SO4. When you pump down, the HF won't be a problem, but you will be concentrating the other acids. Sticking a Teflon tube in through the gas port to drain any liquid in the bottom may be a low tech approach, but it could reduce the need for future maintenance. (If there are any metal salts in there, the evaporation and nitrogen won't get them out.) You might even want to add a quart or so of pure water, then pump it out just to clean out any metal salts.

 

[Hypersphere]

Thanks for your answer and the new angles. Part of the reason I made the thread was that I was a bit unsure as to which questions to ask, so I may in deed have been looking at it the wrong way.

Now the first two things that come to mind are measuring either the volume of contaminants and water in the tank, or checking the humidity. And you might want to run the cooling water at room temperature or above to speed up the process.

Agreed. The humidity and/or dewpoint will be the easiest to measure, so that's what we'll do.

Oh, and again with my engineer hat on, I assume you are checking for leaks while pumping down? If you stop pumping and the pressure starts rising, you either have ice in the container, or a leak.

Well, I'm not a total idiot We mainly manufacture other equipment that work in real vacuums, 10^-6 mbar, so pumping is not that strange to us.

Anyway, my back of the envelope approach would be to run the cooling water while pumping down, and set the pressure limit on the pumping based on the temperature of the cooling water. Use the vapor pressure of water as a guide, though. In a real system that water is going to have plenty of dissolved material.

To stop at ca 20 mbar then. Might be the way to go, especially combined with getting rid of the visible water manually.

Last but not least, have you thought about what concentrated acid may do to the container and its contents? The contaminants are going to include some HF, H2S, H2SO3, and H2SO4. When you pump down, the HF won't be a problem, but you will be concentrating the other acids. Sticking a Teflon tube in through the gas port to drain any liquid in the bottom may be a low tech approach, but it could reduce the need for future maintenance. (If there are any metal salts in there, the evaporation and nitrogen won't get them out.) You might even want to add a quart or so of pure water, then pump it out just to clean out any metal salts.

I'm afraid I have not considered them much. Thanks for sharing that idea and the easy fix though.


On a related note, I was reading a standard relating to the subject, IEEE1125. Apparently the water chemisorbed into the bulk materials takes weeks to extract using vacuum (unknown pressure, might be in a reference). This is obviously impractical from all angles, so it can stay there. Like the standard notes it is reasonable however to measure the dewpoint after a period of 6 months and see if the gas still is fit for operation.